文献速递组学、生物样本库与精准医学医学人工智能与信息学基础与转化研究

不同疾病类型中粪菌移植后菌株定植的变异性与微生物组组成的可预测性

Variability of strain engraftment and predictability of microbiome composition after fecal microbiota transplantation across different diseases.
PubMed 文献 · Nat Med · 2022-09-15

该研究对8种疾病类型的226组供体、移植前受者和移植后受者粪便微生物组进行鸟枪法宏基因组系统Meta分析,发现供体菌株定植较高的受者更易获得临床成功(P=0.017),多途径移植和抗生素治疗的感染性疾病受者定植增加,机器学习预测移植后菌种存在的平均AUROC为0.77。

要点

  • 供体菌株定植较高的受者接受粪菌移植后更可能获得临床成功(P=0.017)。
  • 多途径移植及抗生素治疗的感染性疾病受者定植增加,优于抗生素初治的非传染性疾病患者。
  • 机器学习预测移植后菌种存在的平均AUROC为0.77,可辅助筛选供体。

核心结论

  • 研究揭示了粪菌移植后微生物组定植的物种特异性模式,并展示了能预测供体以优化移植后特定微生物组特征的机器学习模型。

研究设计与样本

  • 对来自8种不同疾病类型的226组供体、移植前受者和移植后受者粪便微生物组进行鸟枪法宏基因组系统Meta分析。

主要发现

  • 在跨研究评估中,供体菌株定植较高的受者接受粪菌移植后更可能获得临床成功(P=0.017)。
  • 在所有队列中,接受多途径粪菌移植(例如同一治疗期间同时通过胶囊和结肠镜)的个体以及抗生素治疗的感染性疾病受者,其定植增加,优于抗生素初治的非传染性疾病患者。
  • 拟杆菌门和放线菌门物种(包括双歧杆菌)的定植高于厚壁菌门,但六种特征不明的厚壁菌门物种除外。
  • 跨数据集机器学习在留一数据集评估中预测移植后受者中物种存在与否的平均AUROC为0.77,并强调了微生物丰度、流行率和分类学对推断移植后物种存在的相关性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

Fecal microbiota transplantation (FMT) is highly effective against recurrent Clostridioides difficile infection and is considered a promising treatment for other microbiome-related disorders, but a comprehensive understanding of microbial engraftment dynamics is lacking, which prevents informed applications of this therapeutic approach. Here, we performed an integrated shotgun metagenomic systematic meta-analysis of new and publicly available stool microbiomes collected from 226 triads of donors, pre-FMT recipients and post-FMT recipients across eight different disease types. By leveraging improved metagenomic strain-profiling to infer strain sharing, we found that recipients with higher donor strain engraftment were more likely to experience clinical success after FMT (P = 0.017) when evaluated across studies. Considering all cohorts, increased engraftment was noted in individuals receiving FMT from multiple routes (for example, both via capsules and colonoscopy during the same treatment) as well as in antibiotic-treated recipients with infectious diseases compared with antibiotic-naïve patients with noncommunicable diseases. Bacteroidetes and Actinobacteria species (including Bifidobacteria) displayed higher engraftment than Firmicutes except for six under-characterized Firmicutes species. Cross-dataset machine learning predicted the presence or absence of species in the post-FMT recipient at 0.77 average AUROC in leave-one-dataset-out evaluation, and highlighted the relevance of microbial abundance, prevalence and taxonomy to infer post-FMT species presence. By exploring the dynamics of microbiome engraftment after FMT and their association with clinical variables, our study uncovered species-specific engraftment patterns and presented machine learning models able to predict donors that might optimize post-FMT specific microbiome characteristics for disease-targeted FMT protocols.

来源:Nat Med,查看原文。版权归原作者所有。

引用

  1. Ianiro G, Punčochář M, Karcher N, et al. Variability of strain engraftment and predictability of microbiome composition after fecal microbiota transplantation across different diseases. Nat Med. 2022. PMID: 36109637. doi:10.1038/s41591-022-01964-3

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。